J. Phys. Chem. A 2009, 113, 3481–3490
3481
Additivity of Substituent Effects. Core-Ionization Energies and Substituent Effects in Fluoromethylbenzenes Thomas X. Carroll,† T. Darrah Thomas,*,‡ Leif J. Sæthre,§ and Knut J. Børve§ DiVision of Natural Sciences, Mathematics, and Physical Education, Keuka College, Keuka Park, New York 14478; Department of Chemistry, Oregon State UniVersity, CorVallis, Oregon 97331; and Department of Chemistry, UniVersity of Bergen, NO-5007 Bergen, Norway ReceiVed: December 2, 2008; ReVised Manuscript ReceiVed: February 12, 2009
Carbon 1s ionization energies have been measured for all of the carbon atoms in eight fluoromethylbenzenes. Enthalpies of protonation have been calculated for protonation at all of the ring carbons in the same molecules. These data together with previously reported data on fluorobenzenes and methylbenzenes provide the basis for studying the additivity of substituent effects and the correlation between enthalpies of protonation with core-ionization energies. Although a linear additivity model accounts reasonably well for both the ionization energies and the enthalpies of protonation, a better description, especially for the enthalpies, is obtained by inclusion of nonlinear terms that account for interactions between two substituents on the same molecule. There are families of nearly parallel correlation lines between enthalpies of protonation and core-ionization energies. The existence of several families can be primarily understood in terms of the linear additivity picture and more completely understood when the nonlinear terms are taken into account. The role of the methyl group as a polarizible π-electron donor is contrasted with the role of fluorine, which is a substituent of low polarizibility that acts to withdraw electrons from the adjacent carbon and to donate electrons through resonance to the ring. The role of the hydrogen atoms as π-electron acceptors in the protonated species is illustrated. Introduction Substituents have a major effect on the properties of benzene and its derivatives. As a result, substituent effects and their systematics have been the subject of many investigations. Carbon 1s photoelectron spectroscopy is potentially an important tool for studying such effects, since it is, in principle, possible to measure the carbon 1s ionization energy for every chemically unique carbon in the benzene ring. Until recently, however, the resolution of the technique has not been adequate to resolve reliably the contributions of the inequivalent carbon atoms to the photoelectron spectra. With the availability of thirdgeneration synchrotrons, high-resolution electron spectrometers, and high-level theoretical calculations, this situation has now changed, and it is now possible to analyze such spectra so as to provide credible information on the various carbon atoms in substituted benzenes. We have recently reported on the carbon 1s photoelectron spectra of fluorobenzenes1 and methylbenzenes.2 In these studies, we were able to assign carbon 1s ionization energies to all of the chemically unique carbon atoms and to show that the experimental results are in good agreement with theoretical predictions. Here, we extend the previous work by considering substituted benzenes having both methyl and fluoro substituents, and, for this purpose, we have measured the carbon 1s photoelectron spectra for eight fluoromethylbenzenes, giving 47 new ionization energies. Core-ionization energies are closely related to other chemical properties that depend on the ability of a molecule to accept charge at a specific site. Among these are acidity, basicity, enthalpies of protonation,3 rates of electrophilic reactions, and * Corresponding author. E-mail:
[email protected]. † Keuka College. ‡ Oregon State University. § University of Bergen.
Hammett parameters. Of particular interest have been correlations between core-ionization energies and enthalpies of protonation, since both of these processes involve addition of a positive charge at a specific site in the molecule. Linear correlations between enthalpies of protonation and core-ionization energies are well established for nitrogen, oxygen, and a few other elements, and these correlations have provided insight into the site of protonation and geometric changes that occur upon protonation.4,5 However, until recently there have not been sufficiently good data available for carbon 1s ionization energies to make investigation of these relationships for carbon possible. We have recently demonstrated that such correlations exist for 1,3-butadiene and 1,3-pentadiene,6 the fluorobenzenes,1 and the methylbenzenes.2 For the fluorobenzenes these correlations provide insight into the π-donating effect of fluorine. The data for the methylbenzenes have extended these correlations and have provided insight into the differences between the π-donating action of the two substituents. For core ionization, we can measure a carbon 1s ionization energy for each carbon atom in the molecule. For protonation this is not the case. In general, experimental enthalpies of protonation are known only for the most favorable site of protonation in each molecule. However, there exist theoretical techniques that allow us to predict these enthalpies for all sites in the molecule,7,8 and, where experimental data are available, there is good agreement between theory and experiment.1 We have used the procedure suggested by Maksic´ et al.8 for calculating the relevant enthalpies (38 energies) and use these theoretically predicted enthalpies for comparison with the coreionization energies. The idea that the effect of two or more substituents on the properties of a benzene ring is simply the sum of their individual effects is referred to as the “additivity principle”. Investigations of this principle in a variety of situations have a long history.8-10
10.1021/jp810612x CCC: $40.75 2009 American Chemical Society Published on Web 03/24/2009
3482 J. Phys. Chem. A, Vol. 113, No. 15, 2009 We have shown that an additivity model that includes only linear terms gives a reasonably good description of the core-ionization energies and enthalpies of protonation of the methyl-2 and fluorobenzenes.1 However, an improved description can be obtained by including nonlinear terms. Here, we extend the previous investigations of the additivity model and establish the relationship between the additivity model and the correlations between enthalpies of protonation and core-ionization energies. The nonlinear terms are of special interest in that they provide information on nonadditive substituent effects. Although these have been much less studied than additive effects, there are numerous examples of chemical properties that differ from those expected from an additivity model.11,12 An example is the dipole moment of nitroaniline, which is significantly larger than the sum of the dipole moments of aniline and nitrobenzene. Recently, a study of substituent interaction energies, combined with a Hammett dual-parameter approach, has shown that field/ inductive effects interact with resonance effects in 1,3- and 1,4disubstituted benzenes.13 A theoretical analysis of intersubstituent interactions in disubstituted benzenes having one donor substituent and one acceptor substituent predicts that the ability of the donor to donate electrons to the ring and the ability of the acceptor to withdraw electrons from the ring are both enhanced if the two substituents are ortho or para to each other and suppressed if they are meta to each other.11 In this study, we have investigated electron-donating and electron-accepting properties of methyl and fluoro substituents, with the object of determining both linear and nonlinear substituent effects. This series of molecules offers a unique possibility to study both of these effects since all sites of the ring carbons can be probed. Procedures and Results Experimental Procedures. Measurements of the carbon 1s photoelectron spectra for eight fluoromethyl-substituted benzene molecules in the gas phase were carried out at beamline I411 of the MAX II synchrotron facility.14 The samples were obtained from commercial sources, and the purity ranged from 97 to 99%. The photon energy was 330 eV. The monochromator slits and the settings of the Scienta SES-200 electron-energy analyzer were chosen to give an overall resolution of about 75 meV. Calibration of the ionization-energy scale and measurement of the actual resolution was made with each compound mixed with carbon dioxide, for which the ionization energies and natural line width are well-known.15,16 The actual resolutions determined in this way ranged from 67 to 96 meV. The spectra were fit by least-squares with peak shapes that include the effects of resolution, core-hole lifetime, vibrational excitation, and the interaction of the photoelectron with the Auger electron that accompanies core ionization (postcollision interaction, or PCI).17 For the effects of PCI and lifetime broadening, we have used the shape given by eq 12 from van der Straten et al.18 This is convoluted with a Gaussian function that represents the instrumental broadening. The carbon dioxide calibration spectrum has three closely spaced peaks, arising from excitation of the symmetric stretching mode in the core-ionized molecule. These spectra are fit constraining the lifetime width, relative positions, and relative heights of the three peaks to previously determined values.15 Thus, the only fitting variables are the instrumental resolution and the peak position for the transition that leaves the molecule with no vibrational excitationsthe adiabatic transition. For
Carroll et al.
Figure 1. Carbon 1s photoelectron spectra of 5-fluoro-m-xylene and 3-fluoro-o-xylene measured at a photon energy of 330 eV. Open circles represent the experimental data. Solid black lines show the overall fit to the data. Colored lines show the contributions from the different carbon atoms.
carbon dioxide this energy is known from previous measurements to be 297.664 eV with an absolute uncertainty of 0.03 eV.16 The fluoromethylbenzene molecules have five to eight chemically inequivalent carbon atoms. This is in contrast to our previous studies, where the maximum number of inequivalent carbon atoms was four for the fluorobenzenes and five for the methylbenzenes. To aid in resolving these spectra, we note that each type of carbon atom produces a unique pattern of vibrational excitation. For each carbon atom we calculate theoretically the vibrational structure associated with core ionization; see below for details. This structure is convoluted with functions that represent the instrumental broadening and the effects of lifetime and postcollision interaction. For the lifetime width we have used 100 meV, which is typical of the lifetime width found for a number of hydrocarbons.19 For the resolution width, we have taken the value given from fitting the carbon dioxide spectrum. The set of profiles are fit to the experimental data by least squares. The fitting parameters are a constant background and the overall intensities and adiabatic (threshold) energies for each carbon. In some cases, it has been necessary to constrain relative intensities to the values expected from the stoichiometry of the molecule. In Figure 1 we present two examples of the experimental carbon 1s photoelectron spectra. These are 5-fluoro-m-xylene (Figure 1a) and 3-fluoro-o-xylene (Figure 1b). For 5-fluoro-mxylene, where there are five chemically different carbon atoms, the fitting is completely straightforward and unambiguous. For 3-fluoro-o-xylene, there are eight chemically distinct carbon atoms, and, in consequence, the fitting presents significant problems. In these figures, the open circles represent the experimental data, the black lines the overall fits, and the colored lines the contributions from the individual peaks. As noted above, the shapes of these individual peaks, which reflect the vibrational profiles, are established by the theoretical calculations. The only
Proton Affinities of Fluoromethylbenzenes free fitting parameters are the positions and intensities of the peaks. Where several peaks are overlapping, their relative intensities have been constrained by the stoichiometry of the molecule. For 5-fluoro-m-xylene, the C5 peak is shifted to higher ionization energy by the fluorine that is attached at this position. In Figure 1a we see this peak at an ionization energy of 292.4 eV, and we see that the shape calculated for this peak agrees well with the observed shape. The shoulder at an ionization energy of 291 eV arises from excitation of the CH stretching mode in the peak resulting from ionization of the methyl carbons. This mode has strong excitation because there are three hydrogens on each methyl group. The presence of this shoulder in the spectrum allows for an unambiguous placement of the methyl peak. The weak shoulder at low ionization energy is assigned to the C2 peak on the grounds that this peak, which is due to a single carbon, will have only half the intensity of the peaks due to the C4/C6 and C1/C3 carbons. The two remaining peaks (C4/C6 and C1/C3) can be placed as shown in Figure 1a or can be interchanged. The order shown in Figure 1a provides a much better fit than the reverse choice and is in agreement with the order predicted by theory. For the fit shown in Figure 1a, we have required that the four peaks contributing to the left-hand structure have relative areas dictated by the stoichiometry of the molecule. If we relax this requirement, we obtain a slightly better fit, but without a significant change in the peak positions, which are within an average of 2 meV of those found from the fits shown in Figure 1a. The relative areas agree with those expected, with a rootmean-square deviation of less than 6%. If we interchange the C4/C6 and C1/C3 peaks, then the resulting fit gives relative intensities that are far from the stoichiometric values. For 3-fluoro-o-xylene, with eight inequivalent carbon atoms, the situation is more complicated. C3 can be uniquely assigned to the small peak at an ionization energy of 292.3 eV, and we see, just as for 5-fluoro-m-xylene, the shape of this peak is predicted well by the theory. The shoulders on the high-energy side of the main peak can, as with 5-fluoro-m-xylene, be associated with the strong vibrational structure from the methyl peak, allowing unambiguous placement of the contribution from ionization of the methyl groups. The two methyl groups are, however, inequivalent, and, consequently we include two peaks. For the remaining five peaks, there are 120 different orders possible for the initial guesses of their positions. Previous experience with m-xylene and fluorobenzene suggests that the ionization energies should be in the order C6 < C4 < C5 < C2 < C1 < CH3. This is also the order indicated by theory and by both an additivity model and correlations with enthalpies of protonation, discussed below. We have tried all 120 possible different initial orderings in the least-squares fits and find that, although there is some variation in the values of χ2, there is no convincing statistical evidence that one is significantly better than another. For lack of a better choice, we have chosen the fit that gives an order that is the same as the order predicted by theory. The corresponding spectrum is shown in Figure 1b. The spectra for the other molecules fall between these extremes. Where the fits do not give unambiguous information on the order of the peaks, we have used the order that agrees with theory. This is the case for three compounds, 3-fluoro-oxylene, 2,5-difluorotoluene, and 2,4-difluorotoluene. Theoretical Procedures. Theoretical calculations include prediction of the vibrational profiles, relative carbon 1s ionization energies, and enthalpies of protonation. For all of these, the Gaussian set of programs was used.20
J. Phys. Chem. A, Vol. 113, No. 15, 2009 3483 Core Ionization. For the calculations related to core ionization, the B3LYP method was used with a triple-ζ basis set plus polarization functions. The core hole was simulated with an effective core potential. Details of the basis set and the effective core potential are given elsewhere21 and in the Supporting Information. For each molecule and for each core-ionized species, we calculate the optimized geometry, vibrational frequencies, and normal modes. Using the harmonic oscillator approximation, we calculate Franck-Condon factors for the excitation of each normal mode in the core-ionized molecule. These sets of Franck-Condon factors are convoluted to produce intensities for the combination modes, and, hence, the full vibrational profile. Typically, each profile includes about 1000 lines, and 50 to 100 lines account for 90% of the intensity. The profile is used in the fitting procedure as described above. Calculations using this method have been found to overestimate the shrinkage of a CH bond that accompanies core ionization of the carbon to which the hydrogen is attached. The overestimate is about 0.3 pm for an sp3 CH bond and about 0.2 pm for an sp2 CH bond.22 The calculated CH bond lengths have, therefore, been adjusted accordingly. For the CH stretching mode of the methyl group, we have found that it is necessary to include the effects of anharmonicity. For this purpose we use Franck-Condon factors calculated using a Morse potential. The vibrational frequencies calculated by this procedure are typically higher than those observed experimentally. On the basis of our experience with other molecules,23 we have scaled all of the calculated frequencies by 0.99 except for those of the CH stretching modes for the hydrogens attached to the core-ionized carbon. In these cases, the scaling factor is 0.95. The electronic structure calculations give energies for the various core-ionized species. Since these involve the approximation of the effective-core potential, they cannot be viewed as absolute energies. However, from the calculated energies of the ionized species, corrected for zero-point energies, we can obtain the predicted ionization energies relative to that of benzene. These are listed in Table 1. Enthalpies of Protonation. We are interested in correlations between enthalpies of protonation and carbon 1s ionization energies. For the latter we have measurements for every position in the molecule, whereas for the former even if measured values are available they are only for the most favorable site of protonation. For a more complete view of protonation, we need to rely on theoretical calculations. For this, we have used the method suggested by Maksic et al.,8 which has been found to give reasonably accurate predictions of enthalpies of protonation. This method involves finding the optimum geometries and zeropoint corrections using the Hartree-Fock method with the 6-31G(d) basis set. With these geometries, the energies are then calculated with the MP2 method and the 6-31G(d,p) basis set. The energies of the molecules are taken to be the MP2 energies plus the zero-point energies corrected by the empirical factor of 0.89.24 Enthalpies of protonation (∆E°0 ) ∆H°0) calculated in this way are listed in Table 1. Results. The results of our experimental measurements and theoretical calculations are given in Table 1 for the eight fluoromethylbenzene molecules that we have investigated. Also included for comparison are the results for benzene.25 The first column of numbers gives the adiabatic ionization energy, which is the energy needed to produce the core-ionized ion in its vibrational ground state. It is obtained directly from the fitting of the experimental data. The second column gives the vertical ionization energy, which is the energy needed to produce the ion in the same geometry as that of the neutral molecule. It is
3484 J. Phys. Chem. A, Vol. 113, No. 15, 2009
Carroll et al.
TABLE 1: Carbon 1s Ionization Energies and Theoretically Calculated Enthalpies of Protonation for the Fluoromethylbenzenes (eV) compound benzene p-fluorotoluene
2,4-difluorotoluene
2,5-difluorotoluene
2,6-difluorotoluene
3,5-difluorotoluene
3-fluoro-o-xylene
2-fluoro-m-xylene
5-fluoro-m-xylene
C1 C2 C3 C4 CH3 C1 C2 C3 C4 C5 C6 CH3 C1 C2 C3 C4 C5 C6 CH3 C1 C2,6 C3,5 C4 CH3 C1 C2,6 C3,5 C4 CH3 C1 C2 C3 C4 C5 C6 CH3 (1) CH3 (2) C1,3 C2 C4,6 C5 CH3 C1,3 C2 C4,6 C5 CH3
adiabatic
vertical
rel adiabatic
rel theory
enthalpy of protonation
290.241 290.425 290.305 290.363 292.485 290.644 290.684 292.802 290.583 292.806 290.452 290.632 290.821 290.850 292.572 290.651 290.620 292.679 290.509 290.863 290.769 292.764 290.444 290.662 290.836 290.927 290.365 292.914 290.546 290.880 290.415 290.301 292.297 290.101 290.202 289.953 290.575 290.535 290.372 292.183 290.077 290.007 290.571 290.462 289.858 290.020 292.415 290.597
290.377 290.577 290.463 290.548 292.570 290.851 290.874 292.892 290.757 292.906 290.668 290.805 291.030 291.025 292.682 290.849 290.825 292.797 290.700 291.065 290.922 292.852 290.657 290.837 290.993 291.085 290.571 293.014 290.735 291.092 290.536 290.444 292.370 290.288 290.363 290.125 290.800 290.762 290.529 292.248 290.236 290.187 290.775 290.597 290.032 290.203 292.497 290.808
0.000 0.184 0.064 0.122 2.244 0.403 0.443 2.561 0.342 2.565 0.211 0.391 0.580 0.609 2.331 0.410 0.379 2.438 0.268 0.622 0.528 2.523 0.203 0.421 0.595 0.686 0.124 2.673 0.305 0.639 0.174 0.060 2.056 -0.140 -0.039 -0.288 0.334 0.294 0.131 1.942 -0.164 -0.234 0.330 0.221 -0.383 -0.221 2.174 0.356
0.000 0.211 0.072 0.107 2.216 0.478 0.475 2.550 0.389 2.573 0.246 0.405 0.653 0.675 2.329 0.453 0.347 2.449 0.325 0.697 0.593 2.522 0.234 0.420 0.676 0.768 0.142 2.689 0.296 0.734 0.141 0.090 2.035 -0.157 -0.081 -0.290 0.397 0.352 0.152 1.891 -0.211 -0.207 0.381 0.246 -0.405 -0.262 2.145 0.422
-7.798 -7.855 -7.768 -7.910 -7.313
obtained from the adiabatic ionization energy by adding the average vibrational excitation energy, which is obtained from the theoretical calculations of the vibrational profile. The third and fourth columns of numbers give the shifts in adiabatic ionization energy relative to benzene as given by experiment and theory. The final column gives the calculated enthalpies of protonation. The uncertainty in the absolute values for the experimental ionization energies is approximately the uncertainty in the ionization energy of the carbon dioxide reference standard, 0.03 eV. Comparisons of the experimental values with the predictions of theory and of an additivity model, discussed below, suggest that the relative uncertainties are less than 0.02 eV and possibly as small as 0.01 eV. The average difference between the experimental and theoretical values (experiment minus theory) of the ionization energies given in Table 1 is -20 meV and the standard deviation is 41 meV. If we consider not only these energies but also the previously reported values for fluorobenzenes and methylben-
-7.841 -7.010 -7.858 -7.020 -7.997 -7.501 -7.462 -7.443 -7.634 -7.834 -7.245 -7.747 -7.737 -7.021 -8.007 -7.555 -7.165 -8.115 -6.794 -8.049 -7.761 -8.024 -7.439 -8.214 -7.949 -8.269 -7.892 -7.644 -8.094 -8.161 -7.606 -8.410 -8.347 -7.247
zenes (85 energies in all), then the corresponding values are -14 and 42 meV. Closer examination of the results indicates that this comparison is misleading as to the extent of disagreement between experiment and theory. In earlier papers1,2,6,26 we noted that the theoretical method that we have used overestimates the substituent-induced shifts. This overestimation appears to arise from the use of the B3LYP method. Calculations based on the ccpVTZ basis set and the CCSD(T) method for 66 different carbon atoms in 25 different compounds lead to underestimates of the shifts by about 3%.27 For the same set of compounds, the B3LYP method with the cc-pVTZ basis set overestimates the shifts by 6%. Although the CCSD(T) procedure gives a better prediction, it is not easily used for the highly substituted benzene molecules considered here. If we consider all of the data for the substituted benzenes, we find that the shifts predicted by the B3LYP procedure are overestimated by about 6% for a CF carbon (a carbon atom
Proton Affinities of Fluoromethylbenzenes
J. Phys. Chem. A, Vol. 113, No. 15, 2009 3485
bonded to fluorine) and by about 9% for other carbon atoms. Allowing for this effect, we find that this theory accounts for the observed shifts with a standard deviation of 16 meV for the first class of carbon atoms and 21 meV for the second class. This generally good agreement between the experimentally determined shifts and the theoretically calculated shifts indicates that the methods that we have used to extract the experimental values from the spectra are approximately correct. Also, as noted above, this agreement provides us with an estimate that the uncertainties in the experimental values are no more than about 20 meV. Additivity of Substituent Effects The effects of multiple substituents on benzene have often been described in terms of an additivity model. That is, the total effect of the substituents is considered to be the sum of independent effects of the individual substituents. A more complete model allows for the possibility of quadratic or cross terms. Thus, we can write
∆Er )
∑ nkRkr+ ∑ (nk(nk - 1)βkr/2 + ∑ nknlγklr) k
k
l>k
where ∆Er is the energy change for the reaction of interest relative to that of some reference compound (benzene, in this case). The indices k and l identify the type of substituent and its location. Thus, for the case at hand, we have methyl substituents that can be ipso (i), ortho (o), meta (m), or para (p) to the site of interest and fluoro substituents that can be similarly located, identified as i′, o′, m′, and p′. The coefficients R, β, and γ are characteristic of the substituent, its location, and the reaction. The term that involves the coefficient R represents the linear additive effect of the substituents. The remaining terms reflect, to first approximation, departures from simple additivity. The term involving β is nonzero only for ortho and meta substituents; for the fluoromethylbenzenes there are potentially four such terms, oo, mm, o′o′, and m′m′.28 For the summation over γ, there are potentially 26 terms, six involving pairs of methyl groups (io, im, ip, om, op, mp), 6 similar terms for the fluoro groups, and 14 involving one methyl group and one fluoro group. In all, there are 38 terms. Using either measured adiabatic or theoretical ionizationenergy shifts or calculated enthalpies of protonation, we have used least-squares methods to determine the coefficients for the effects of methyl and fluoro substituents on the core-ionization energies and enthalpies of protonation of substituted benzenes. We have used two different approaches. In the first, we consider only the linear terms; in the second, we include also the quadratic terms. Additivity Model for Ionization Energies. With only linear parameters, the additivity model accounts well for the experimental ionization energies with an rms deviation of 0.026 eV. The value of R2 for the correlation is 0.9996, indicating that linear additivity alone provides a quite satisfactory description of the shifts. The linear model works equally well for the theoretically calculated shifts (rms ) 0.025 eV, R2 ) 0.9997). Although the linear additivity model gives a good description of the ionization-energy shifts, inspection of the results shows a few striking discrepancies between the actual values of the shifts and those predicted by the additivity model. For instance, the ionization-energy shifts predicted for 1,2,4,5-tetramethylbenzene are too high by 0.076 eV (C1,2,4,5) and 0.063 eV (C3,6). This effect is seen in both the experimentally measured
TABLE 2: Additivity Coefficients for Methyl and Fluoro Substituents (eV) carbon 1s ionization energies CH3 i CH3 o CH3 m CH3 p F i′ F o′ F m′ F p′ io oma op mp i′o′ i′p′ o′(o′ - 1) o′m′a m′p′ i′o i′m i′p io′ im′ oo′ om′a mm′ pm′
number
theory
expt
protonation
15 24 24 15 24 36 36 24 4 6 5 4 10 8 12 18 10 5 4 2 5 4 4 7 4 5
0.055(5) -0.280(4) -0.170(3) -0.264(6) 2.499(4) 0.281(3) 0.365(3) 0.139(4) 0.044(9) 0.025(4) 0.019(7)
0.044(6) -0.262(5) -0.159(4) -0.227(8) 2.496(5) 0.256(3) 0.339(4) 0.121(5) 0.051(12) 0.024(6) 0.027(10)
0.039(4) -0.268(3) -0.136(3) -0.329(6) 0.853(5) 0.018(3) 0.312(3) -0.080(3)
-0.020(3)
-0.011(4)
-0.027(7)
-0.019(9)
-0.024(6)
-0.0015(8)
-0.022(7)
-0.017(9)
0.019(3) 0.025(6) 0.038(7) -0.115(4) -0.061(6) 0.032(6) -0.025(4) -0.085(6) -0.016(6) -0.035(10) -0.021(5) -0.015(5) 0.010(6) -0.025(4) -0.014(6) -0.024(5)
a
A pair of substituents at the ortho and meta positions can be either ortho to each other or para to each other. We see some evidence that these two possibilities should be treated separately, but since the ortho-meta coefficients are small we do not pursue this question here.
and the theoretically predicted shifts. Similarly, for 1,3difluorobenzene and 1,3,5-trifluorobenzene, the shifts predicted for the CF carbons are too high by 0.047 and 0.071 eV. It is useful, therefore, to consider the possible contributions from the nonlinear terms. A fit to the data using all 38 possible terms gives excellent agreement between observation and prediction, but for 6 of the nonlinear parameters the uncertainty in the parameter is larger than its value and for a number of the nonlinear parameters the statistical significance is small. We find that we can get a good description of the results by including only 7 of the nonlinear parameters. These are io (two methyl groups, one at the ipso position relative to the site of ionization and one at the ortho position), om, op, o′m′ (two fluoro groups, one ipso, and one meta), oi′, om′, and pm′. The values of the parameters determined in this way are listed in Table 2. Also shown as the first column of numbers in Table 2 is the number of data points that contribute to the determination of each coefficient. The experimentally determined shifts differ from those predicted by this model by an rms deviation of 0.019 eV, with R2 ) 0.9998. The corresponding values for the theoretically calculated shifts are 0.014 eV and 0.99990. Adding additional nonlinear parameters to the correlation makes only small improvements in these numbers. (The difference between the rms values for theory and experiment presumably arises from experimental errors in the measurements. The square root of the difference between the squares of these two values provides an estimate of 0.013 eV for the experimental uncertainty for the relative ionization energies.) Of the seven nonlinear parameters, six involve at least one methyl group, and for five of these the methyl group is ortho to
3486 J. Phys. Chem. A, Vol. 113, No. 15, 2009 the site of ionization. Only one of the nonlinear terms involves two fluoro substituents. These results can be understood, at least qualitatively, in terms of the polarizibility of the methyl group. When a polarizible group is added to a molecule, it tends to be polarized in such a direction as to counteract whatever effects were produced by the already existing substituents.29,30 In support of this view we note that the nonlinear terms om, op, oi′, om′, and pm′ all have signs that are opposite to that of the corresponding linear terms: m, p, i′, and m′. This effect is less apparent for fluoro substituents owing to the lower polarizibility of fluorine relative to methyl. Additivity and Enthalpies of Protonation. With only linear terms the additivity model describes 85 enthalpies of protonation with an rms deviation of 0.046 eV (4.5 kJ/mol) and R2 ) 0.993. This restricted model does not describe the enthalpies of protonation as well as it describes the core-ionization energies (0.026 eV and 0.9996, as noted above). If we use the full set of 38 parameters we have an rms deviation of 0.0097 eV with R2 ) 0.9998, but 12 of the parameters are not statistically significant and two others are small enough to be neglected (